You're glossing over the fact that Watson knew the correct answer most of the time. That's what this is about, and that's why this is a significant result.
It looked like Watson had the correct response about 75% of the time. That's probably better than most normal season contestants and home viewers, but I am highly confident that Ken and Brad knew significantly more than 75% of the correct responses.
Doesn't matter, you're discussing a different set of goals. Which might make an interesting game, but ignores the purpose of this exercise. The question on everyone's mind was "can a machine beat humans at Jeopardy". The answer appears to be: yes, very much so.
Whether or not we can make a more interesting or "fair" game of Jeopardy involving such a machine is an entirely separate, and to my mind far less interesting, question.
Where do we go from here? It would seem silly to say "we make a more interesting Humans vs. machines Jeopardy game." Rather, it seems more prudent to figure out ways to expand on this research and use the underlying technology to solve more interesting, and practical, problems.
